PennyLane Quantum Machine Learning Demos

Over 50 runnable Jupyter notebooks covering every major topic in quantum machine learning and variational quantum algorithms. Each demo is a self-contained tutorial written by Xanadu researchers or community contributors.

The PennyLane QML demos are not a linear course but a curated collection of high-quality applied tutorials. Each covers a specific algorithm, technique, or research result with a complete PennyLane implementation. The collection spans from foundational variational methods through cutting-edge research topics, making it useful for both learners filling in gaps and researchers looking for reference implementations.

What you’ll learn

• The variational quantum eigensolver (VQE): the algorithm for estimating ground state energies of molecular Hamiltonians, ansatz circuit design, and optimisation landscape
• The quantum approximate optimisation algorithm (QAOA): problem encoding into Ising models, the QAOA circuit structure, and hyperparameter optimisation
• Quantum kernels and kernel methods: how to use quantum feature maps to compute kernel matrices and feed them into support vector machines and other classical methods
• Quantum neural networks: parameterised circuit architectures, training with gradient descent, and the connection to classical neural networks
• Barren plateaus: the problem of exponentially vanishing gradients in deep quantum circuits, why it is a fundamental obstacle, and mitigation strategies
• Quantum graph neural networks: how to encode graph-structured data into quantum circuits and the potential advantages for graph learning problems
• Differentiable quantum computing: the parameter-shift rule for computing exact gradients of quantum circuits and how PennyLane implements automatic differentiation

Course structure

The demos are organised by category on the PennyLane website: algorithms, quantum machine learning, optimization, and devices. Within each category, tutorials range from introductory to research-level. Each demo is a standalone Jupyter notebook that can be run locally or opened in a hosted environment.

Every notebook follows a consistent pattern: motivation and background, mathematical formulation, PennyLane implementation, results and visualisation, and pointers to related work and papers. Many demos include links to the original research paper they implement.

The collection is actively maintained and grows as new QML research emerges. Demos covering recent results from major conferences typically appear within months of publication.

Who is this for?

• Learners who have basic quantum circuit knowledge and want applied QML content
• Researchers who want a working PennyLane implementation of a specific algorithm
• ML engineers exploring the intersection of machine learning and quantum computing
• Quantum computing practitioners who want to stay current with QML research through practical implementations

Prerequisites

Comfortable Python and NumPy skills are required. Familiarity with quantum circuits (qubits, gates, measurement) at the level of the Xanadu Codebook or equivalent is assumed. For the machine learning demos, familiarity with gradient descent optimisation and basic ML concepts is needed. The demos vary in difficulty; many include explanatory sections that introduce domain-specific prerequisites.

Hands-on practice

Every demo is a working Jupyter notebook. Across the collection you will:

• Run VQE for small molecules (H2, LiH) and compare to exact diagonalisation
• Implement QAOA for the Max-Cut problem on small graphs and sweep over circuit depth
• Compute quantum kernel matrices and train a quantum SVM on a classification dataset
• Build and train a quantum neural network on a toy dataset and observe barren plateau effects
• Implement the parameter-shift rule from scratch and verify it against finite differences
• Modify existing demos to test variations: different ansatze, optimisers, or problem instances

Notebooks can be executed locally with a PennyLane installation or opened in Google Colab with no local setup.

Why take this course?

The PennyLane demos are the best available resource for anyone who wants to see quantum machine learning concepts implemented correctly and completely. Rather than toy examples that only demonstrate that code runs, these demos implement real algorithms with proper analysis of results.

The research-to-demo pipeline at Xanadu means that cutting-edge topics are covered quickly and accurately. For a researcher who wants to understand what barren plateaus look like in practice, or how a quantum graph neural network is actually structured, these demos provide reference implementations that would otherwise require significant effort to build from scratch.

The free, open-source format means every notebook can be inspected, modified, and extended. This makes the collection valuable not just as learning material but as a starting point for original quantum machine learning research.